Platform would have to be reaaaaaaaally still for free space optics to work.

Internet over HF radio is feasible.

• Pros:

1. extremely long distance,
2. over the horizon capability,
3. no line of sight restriction
4. Well-developed radio equipment
5. Not too expensive - mil surplus available.

• Cons:

1. very low bandwidth (physical constraint of signaling at those frequency ranges) 
2. susceptible to weather interference.
3. Best for text-only coms
4. Can be jammed
5. requires another gateway somwhere on the other end attached to the internet
6. requires a dedicated send and dedicated receive frequency
7. It's a point-to-point or point-to-multipoint connection, but a robust mesh topology requires multiples of all equipment and frequencies

I worked with the Battle Force Email system a bit back when I was in the military.

• satellite coverage in certain areas of the world can be quite expensive, it's not like Dish network back in the states. We pay tens of thousands of dollars per month for fractional T1 access at various sites in Iraq. The satellite footprint for internet just may not be there, depending on where you set up shop- the empty ocean has not been a huge target market for satellite content providers so far.

Satellite services on the ocean are likely(correct me if I am wrong) geared toward craft moving with some regularity.  Would this mean those services tend to have a decent distribution of satellites to handle that movement and keep the ships in touch with the net with more regularity?  If this is the case, the footprint may be larger than we suspect and would potentially provide a semi-mobile seastead with better quality internet than ships using the same service.

I did some cursory research on marine internet a while back. It is very expensive. The problem is not the downlink. There are high power Ku band satellites, like the ones used for DirectPC, over much of the globe. Keeping the dish pointed at the satellite even on a moving ship is not technically difficult, and there are a number of vendors with off the shelf solutions. The problem is the UPLINK. Contrary to what many people think, the uplink data on satellite internet doesn't go through the satellite. It goes through some terrestial wireless network. In the case of marine internet it goes over a marine band radio data network. So you wind up with an assymetrical link with slow uploads and a very expensive price.

There currently isn't any better solution, so seasteaders who wish to live way out from shore outside of any nation's EEZ (200 nm or more) will have to either develop a new system or deal with the price and performance of existing systems. For anything less than a global community of seasteaders, satellites are not an option. They are simply too expensive. Your best bet is probably a com platform on a high altitude balloon with a dedicated high speed link to the mainland. Developing such a system may cost nearly as much as developing the seastead itself.

• http://www.dbcde.gov.au/__data/assets/pdf_file/0011/83819/coverage_maps.pdf
• http://www.idirect.net/page.ww?section=News&name=View+News+Release&pressrelease.id=261
• http://www.satsig.net/ivsat-asia.htm

You would definitely want a caching proxy server, no matter what. HF is not a great format for web browsing. I can be done if the site you're browsing doesn't use file that are so big they time out downloading. But it's primarily a text-based system. It's excellent for data like weather, text-email, GPS, etc. Theoretically, yes, you can run different services through different gateways- DNS requests over HF AX.25, or even HTTP requests which then download from Satellite. The implementation is a little tricker. You're not going to beat satellite latency that way though.

• HF is far lower bandwidth than the high frequencies used by satellite
• You're still bouncing a light-speed signal off the ionosphere and around th world to a ground station. Not quite as far as a satellite, but not close either
• You're going to get more re-transmits due to signal/noise ratio

Heath, caching sites such as wikipedia is probably not the best way to go.  Its simply too unlikely most of it will ever be used.  Though caching its top few thousand pages may be a fair idea.  Bastidge is right on using a caching proxy.  My own suggestion on such a proxy would be to extend the cached websites TTL and also let the servers query cached sites for updates during hours that average the lowest bandwidth usage.  This way the most commonly accessed sites are kept nice and up to date, DNS poisoning is less likely to happen(I'm paranoid, sue me :P ), and the updates will have the lowest perceivable impact on the network.

It may also be worth using(as an auxillary not replacement) a large lan or even man network(depending on community size) to create an "internet" by and for that community to host information relevant to the community.  A plus is that it'd work faster, a downside is it would be limited in access.  So I think a combination of the two would be useful.  But that's just me.

I've made some progress on this.

http://wiki.seasteading.org/index.php/Heath/FX.25